lm-proxy

KO | EN

gitlite — search

#javascript #python #hacktoberfest #golang #go #rust #deep-learning #nodejs #macos #awesome #awesome-list #security #mcp #open-source

(★ 150)

OpenAI-compatible HTTP LLM proxy / gateway for multi-provider inference (Google, Anthropic, OpenAI, PyTorch). Lightweight, extensible Python/FastAPI—use as library or standalone service.

File Explorer

Download Latest Version (.zip)

- - badge-publish.yml
  - code-style.yml
  - docker.yml
  - gito-code-review.yml
  - gito-react-to-comments.yml
  - tests.reusable.yml
  - tests.yml
- config.toml
- fallback.md
- http_headers.md
- https.md
- load_balancer_config.py
- vertex-ai.toml
- - __init__.py
  - allow_all.py
  - in_config.py
  - with_request.py
- - __init__.py
  - json.py
  - python.py
  - toml.py
  - yaml.py
- - __init__.py
  - forward_http_headers.py
  - rate_limiter.py
- - __init__.py
  - fallback.py
- __init__.py
- __main__.py
- app.py
- base_types.py
- bootstrap.py
- config.py
- core.py
- errors.py
- loggers.py
- models_endpoint.py
- utils.py
- - lm-proxy_1_hacker_1600x672.png
- - __init__.py
  - config_fn.py
  - extra_headers.yml
  - no_api_key_check.yml
  - rate_limiter.yml
  - test_config.json
  - test_config.toml
  - test_config.yml
  - test_disabled.py
- __init__.py
- conftest.py
- test_api_key_check_allow_all.py
- test_api_key_check_errors.py
- test_api_key_check_with_request.py
- test_config_loaders.py
- test_integration.py
- test_integration_headers.py
- test_loggers.py
- test_models_endpoint.py
- test_print_stream.py
- test_rate_limiter_integration.py
- test_resolve_connection_and_model.py
- test_utils.py
.dockerignore
.env.template
.flake8
.gitignore
AGENTS.md
CLAUDE.md
config.toml
coverage.svg
Dockerfile
LICENSE
Makefile
multi-build.py
poetry.lock
pyproject.toml
README.md

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

<h1 align="center"><a href="#">LM-Proxy</a></h1>

 Lightweight, OpenAI-compatible HTTP proxy server / gateway unifying access to multiple Large Language Model providers and local inference through a single, standardized API endpoint. 


 <a href="https://pypi.org/project/lm-proxy/"><img src="https://img.shields.io/pypi/v/lm-proxy?color=blue" alt="PyPI"></a>
 <a href="https://github.com/Nayjest/lm-proxy/actions/workflows/tests.yml"><img src="https://github.com/Nayjest/lm-proxy/actions/workflows/tests.yml/badge.svg" alt="Tests"></a>
 <a href="https://github.com/Nayjest/lm-proxy/actions/workflows/code-style.yml"><img src="https://github.com/Nayjest/lm-proxy/actions/workflows/code-style.yml/badge.svg" alt="Code Style"></a>
 <img src="https://raw.githubusercontent.com/Nayjest/lm-proxy/coverage-badge/coverage.svg" alt="Code Coverage">
 <a href="https://www.bestpractices.dev/projects/11364"><img src="https://www.bestpractices.dev/projects/11364/badge"></a>
 
 <a href="https://github.com/vshymanskyy/StandWithUkraine/blob/main/README.md"><img src="https://raw.githubusercontent.com/vshymanskyy/StandWithUkraine/refs/heads/main/badges/StandWithUkraine.svg" alt="Stand With Ukraine"></a>
 <a href="https://github.com/Nayjest/lm-proxy/blob/main/LICENSE"><img src="https://img.shields.io/github/license/Nayjest/lm-proxy?color=d08aff" alt="License"></a>

Built with Python, FastAPI and [MicroCore](https://github.com/Nayjest/ai-microcore), **LM-Proxy** seamlessly integrates cloud providers like Google, Anthropic, and OpenAI, as well as local PyTorch-based inference, while maintaining full compatibility with OpenAI's API format.

It works as a drop-in replacement for OpenAI's API, allowing you to switch between cloud providers and local models without modifying your existing client code.

**LM-Proxy** supports **real-time token streaming**, **secure Virtual API key management**, and can be used both as an importable Python library and as a standalone HTTP service. Whether you're building production applications or experimenting with different models, LM-Proxy eliminates integration complexity and keeps your codebase **provider-agnostic**.

## Table of Contents
- [Overview](#lm-proxy)
- [Features](#-features)
- [Getting Started](#-getting-started)
  - [Installation](#installation)
  - [Quick Start](#quick-start)
  - [Run with Docker](#run-with-docker)
- [Configuration](#-configuration)
  - [Basic Structure](#basic-structure)
  - [Environment Variables](#environment-variables)
- [Proxy API Keys vs. Provider API Keys](#-proxy-api-keys-vs-provider-api-keys)
- [API Usage](#-api-usage)
  - [Chat Completions Endpoint](#chat-completions-endpoint)
  - [Models List Endpoint](#models-list-endpoint)
- [User Groups Configuration](#-user-groups-configuration)
  - [Basic Group Definition](#basic-group-definition)
  - [Group-based Access Control](#group-based-access-control)
  - [Connection Restrictions](#connection-restrictions)
  - [Virtual API Key Validation](#virtual-api-key-validation)
- [Advanced Usage](#%EF%B8%8F-advanced-usage)
  - [Dynamic Model Routing](#dynamic-model-routing)
  - [Load Balancing Example](#load-balancing-example)
  - [Google Vertex AI Example](#google-vertex-ai-configuration-example)
  - [Using Tokens from OIDC Provider as Virtual/Client API Keys](#using-tokens-from-oidc-provider-as-virtualclient-api-keys)
- [Add-on Components](#-add-on-components)
  - [Database Connector](#database-connector)
- [Request Handlers (Middleware)](#-request-handlers--middleware)
- [Guides & Reference](#-guides--reference)
- [Known Limitations](#-known-limitations)
- [Debugging](#-debugging)
  - [Debugging Mode](#debugging-mode)
  - [Printing LLM Requests and Responses](#printing-llm-requests-and-responses)
- [Contributing](#-contributing)
- [License](#-license)

## ✨ Features<a id="-features"></a>

- **Provider Agnostic**: Connect to OpenAI, Anthropic, Google AI, local models, and more using a single API
- **Unified Interface**: Access all models through the standard OpenAI API format
- **Dynamic Routing**: Route requests to different LLM providers based on model name patterns
- **Stream Support**: Full streaming support for real-time responses
- **API Key Management**: Configurable API key validation and access control
- **Easy Configuration**: Simple TOML/YAML/JSON/Python configuration files for setup
- **Extensible by Design**: Minimal core with clearly defined extension points, enabling seamless customization and expansion without modifying the core system.

## 🚀 Getting Started<a id="-getting-started"></a>

### Requirements
Python 3.11 | 3.12 | 3.13

### Installation<a id="installation"></a>
```bash
pip install lm-proxy
```
For proxying to Anthropic API or Google Gemini via Vertex AI or Google AI Studio, install optional dependencies:
```
pip install lm-proxy[anthropic,google]
```
or
```
pip install lm-proxy[all]
```

### Quick Start<a id="quick-start"></a>

#### 1. Create a `config.toml` file:
```toml
host = "0.0.0.0"
port = 8000

[connections]
[connections.openai]
api_type = "open_ai"
api_base = "https://api.openai.com/v1/"
api_key = "env:OPENAI_API_KEY"

[connections.anthropic]
api_type = "anthropic"
api_key = "env:ANTHROPIC_API_KEY"

[routing]
"gpt*" = "openai.*"
"claude*" = "anthropic.*"
"*" = "openai.gpt-3.5-turbo"

[groups.default]
api_keys = ["YOUR_API_KEY_HERE"]
```
> **Note** ℹ️
> To enhance security, consider storing upstream API keys in operating system environment variables rather than embedding them directly in the configuration file. You can reference these variables in the configuration using the env:<VAR_NAME> syntax.

#### 2. Start the server:
```bash
lm-proxy
```
Alternatively, run it as a Python module:
```bash
python -m lm_proxy
```

#### 3. Use it with any OpenAI-compatible client:
```python
from openai import OpenAI

client = OpenAI(
    api_key="YOUR_API_KEY_HERE",
    base_url="http://localhost:8000/v1"
)

completion = client.chat.completions.create(
    model="gpt-5",  # This will be routed to OpenAI based on config
    messages=[{"role": "user", "content": "Hello, world!"}]
)
print(completion.choices[0].message.content)
```

Or use the same endpoint with Claude models:
```python
completion = client.chat.completions.create(
    model="claude-opus-4-1-20250805",  # This will be routed to Anthropic based on config
    messages=[{"role": "user", "content": "Hello, world!"}]
)
```

### Run with Docker<a id="run-with-docker"></a>

Official images are published to GitHub Container Registry on every release
(`latest`, `3`, `3.3`, `3.3.0`, …; `edge` tracks the `main` branch).
They include the Anthropic and Google connectors and YAML config support out of the box.

Mount your `config.toml` and provide API keys via an `.env` file:

```bash
docker run -d --name lm-proxy \
    -p 8000:8000 \
    -v ./config.toml:/app/config.toml:ro \
    --env-file .env \
    lmproxy/lmproxy:edge
```

Any value referenced in the config as `env:<VAR_NAME>` (upstream provider keys,
client API keys, etc.) can be defined in the `.env` file:

```ini
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
```

See [examples/docker-compose.yml](examples/docker-compose.yml) for a Docker Compose setup.

To bake additional Python packages into the image (e.g. for
[database log storage](#database-connector)), build it with the `EXTRA_PIP_PACKAGES` argument:

```bash
docker build --build-arg EXTRA_PIP_PACKAGES="lm-proxy-db-connector sqlalchemy psycopg2-binary" -t lm-proxy .
```

## 📝 Configuration<a id="-configuration"></a>

LM-Proxy is configured through a TOML/YAML/JSON/Python file that specifies connections, routing rules, and access control.

### Basic Structure<a id="basic-structure"></a>
```toml
host = "0.0.0.0" # Interface to bind to
port = 8000 # Port to listen on
dev_autoreload = false # Enable for development
print_stream = false # Print LLM requests / responses to stdout, see "Debugging"

# API key validation function (optional)
api_key_check = "lm_proxy.api_key_check.check_api_key_in_config"

# LLM Provider Connections
[connections]

[connections.openai]
api_type = "open_ai"
api_base = "https://api.openai.com/v1/"
api_key = "env:OPENAI_API_KEY"

[connections.google]
api_type = "google"
api_key = "env:GOOGLE_API_KEY"

[connections.anthropic]
api_type = "anthropic"
api_key  = "env:ANTHROPIC_API_KEY"

# Routing rules (model_pattern = "connection.model")
[routing]
"gpt*" = "openai.*"     # Route all GPT models to OpenAI
"claude*" = "anthropic.*"  # Route all Claude models to Anthropic
"gemini*" = "google.*"  # Route all Gemini models to Google
"*" = "openai.gpt-3.5-turbo"  # Default fallback

# Access control groups
[groups.default]
api_keys = [
    "KEY1",
    "KEY2"
]

# optional
[[loggers]]
class = 'lm_proxy.loggers.BaseLogger'
[loggers.log_writer]
class = 'lm_proxy.loggers.JsonLogWriter'
file_name = 'storage/json.log'
[loggers.entry_transformer]
class = 'lm_proxy.loggers.LogEntryTransformer'
completion_tokens = "response.usage.completion_tokens"
prompt_tokens = "response.usage.prompt_tokens"
prompt = "request.messages"
response = "response"
group = "group"
connection = "connection"
api_key_id = "api_key_id"
remote_addr = "remote_addr"
created_at = "created_at"
duration = "duration"
```

### Environment Variables<a id="environment-variables"></a>

You can reference environment variables in your configuration file by prefixing values with `env:`.

For example:
```toml
[connections.openai]
api_key = "env:OPENAI_API_KEY"
```

At runtime, LM-Proxy automatically retrieves the value of the target variable
(OPENAI_API_KEY) from your operating system's environment or from a .env file, if present.

### .env Files

By default, LM-Proxy looks for a `.env` file in the current working directory
and loads environment variables from it.

You can refer to the [.env.template](https://github.com/Nayjest/lm-proxy/blob/main/.env.template)
 file for an example:
```dotenv
OPENAI_API_KEY=sk-u........
GOOGLE_API_KEY=AI........
ANTHROPIC_API_KEY=sk-ant-api03--vE........

# "1", "TRUE", "YES", "ON", "ENABLED", "Y", "+" are true, case-insensitive.
# See https://github.com/Nayjest/ai-microcore/blob/v4.4.3/microcore/configuration.py#L36
LM_PROXY_DEBUG=no
```

You can also control `.env` file usage with the `--env` command-line option:
```bash
# Use a custom .env file path
lm-proxy --env="path/to/your/.env"
# Disable .env loading
lm-proxy --env=""
```

## 🔑 Proxy API Keys vs. Provider API Keys<a id="-proxy-api-keys-vs-provider-api-keys"></a>

LM-Proxy utilizes two distinct types of API keys to facilitate secure and efficient request handling.

- **Proxy API Key (Virtual API Key, Client API Key):**  
A unique key generated and managed within LM-Proxy.  
Clients use these keys to authenticate their requests to the proxy's API endpoints.  
Each Client API Key is associated with a specific group, which defines the scope of access and permissions for the client's requests.  
These keys allow users to securely interact with the proxy without direct access to external service credentials.

- **Provider API Key (Upstream API Key):**
A key provided by external LLM inference providers (e.g., OpenAI, Anthropic, Mistral, etc.) and configured within the LM-Proxy.  
The proxy uses these keys to authenticate and forward validated client requests to the respective external services.  
Provider API Keys remain hidden from end users, ensuring secure and transparent communication with provider APIs.

This distinction ensures a clear separation of concerns: 
Virtual API Keys manage user authentication and access within the proxy, 
while Upstream API Keys handle secure communication with external providers.

## 🔌 API Usage<a id="-api-usage"></a>

LM-Proxy implements the OpenAI chat completions API endpoint. You can use any OpenAI-compatible client to interact with it.

### Chat Completions Endpoint<a id="chat-completions-endpoint"></a>
```http
POST /v1/chat/completions
```

#### Request Format
```json
{
  "model": "gpt-3.5-turbo",
  "messages": [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "What is the capital of France?"}
  ],
  "temperature": 0.7,
  "stream": false
}
```

#### Response Format
```json
{
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "The capital of France is Paris."
      },
      "finish_reason": "stop"
    }
  ]
}
```

### Models List Endpoint<a id="models-list-endpoint"></a>

List and describe all models available through the API.
```http
GET /v1/models
```

The **LM-Proxy** dynamically builds the models list based on routing rules defined in `config.routing`.  
Routing keys can reference both **exact model names** and **model name patterns** (e.g., `"gpt*"`, `"claude*"`, etc.).

By default, wildcard patterns are displayed as-is in the models list (e.g., `"gpt*"`, `"claude*"`).  
This behavior can be customized via the `model_listing_mode` configuration option:
```
model_listing_mode = "as_is" | "ignore_wildcards" | "expand_wildcards"
```

Available modes:

- **`as_is`** *(default)* — Lists all entries exactly as defined in the routing configuration, including wildcard patterns.  
- **`ignore_wildcards`** — Excludes wildcard patterns, showing only explicitly defined model names.  
- **`expand_wildcards`** — Expands wildcard patterns by querying each connected backend for available models *(feature not yet implemented)*.

To obtain a complete and accurate model list in the current implementation,
all supported models must be explicitly defined in the routing configuration, for example:
```toml
[routing]
"gpt-4" = "my_openai_connection.*"
"gpt-5" = "my_openai_connection.*"
"gpt-8"= "my_openai_connection.gpt-3.5-turbo"
"claude-4.5-sonnet" = "my_anthropic_connection.claude-sonnet-4-5-20250929"
"claude-4.1-opus" = "my_anthropic_connection.claude-opus-4-1-20250805"
[connections]
[connections.my_openai_connection]
api_type = "open_ai"
api_base = "https://api.openai.com/v1/"
api_key  = "env:OPENAI_API_KEY"
[connections.my_anthropic_connection]
api_type = "anthropic"
api_key  = "env:ANTHROPIC_API_KEY"
```

#### Response Format
```json
{
  "object": "list",
  "data": [
    {
      "id": "gpt-6",
      "object": "model",
      "created": 1686935002,
      "owned_by": "organization-owner"
    },
    {
      "id": "claude-5-sonnet",
      "object": "model",
      "created": 1686935002,
      "owned_by": "organization-owner"
    }
  ]
}
```

## 🔒 User Groups Configuration<a id="-user-groups-configuration"></a>

The `[groups]` section in the configuration defines access control rules for different user groups.  
Each group can have its own set of virtual API keys and permitted connections.

### Basic Group Definition<a id="basic-group-definition"></a>
```toml
[groups.default]
api_keys = ["KEY1", "KEY2"]
allowed_connections = "*" # Allow access to all connections
```

### Group-based Access Control<a id="group-based-access-control"></a>

You can create multiple groups to segment your users and control their access:
```toml
# Admin group with full access
[groups.admin]
api_keys = ["ADMIN_KEY_1", "ADMIN_KEY_2"]
allowed_connections = "*"  # Access to all connections

# Regular users with limited access
[groups.users]
api_keys = ["USER_KEY_1", "USER_KEY_2"]
allowed_connections = "openai,anthropic"  # Only allowed to use specific connections

# Free tier with minimal access
[groups.free]
api_keys = ["FREE_KEY_1", "FREE_KEY_2"]
allowed_connections = "openai"  # Only allowed to use OpenAI connection
```

### Connection Restrictions<a id="connection-restrictions"></a>

The `allowed_connections` parameter controls which upstream providers a group can access:

- `"*"` - Group can use all configured connections
- `"openai,anthropic"` - Comma-separated list of specific connections the group can use

This allows fine-grained control over which users can access which AI providers, enabling features like:

- Restricting expensive models to premium users
- Creating specialized access tiers for different user groups
- Implementing usage quotas per group
- Billing and cost allocation by user group

### Virtual API Key Validation<a id="virtual-api-key-validation"></a>

#### Overview

LM-Proxy includes 2 built-in methods for validating Virtual API keys:
 - `lm_proxy.api_key_check.check_api_key_in_config` - verifies API keys against those defined in the config file; used by default
 - `lm_proxy.api_key_check.CheckAPIKeyWithRequest` - validates API keys via an external HTTP service

The API key check method can be configured using the `api_key_check` configuration key.  
Its value can be either a reference to a Python function in the format `my_module.sub_module1.sub_module2.fn_name`,
or an object containing parameters for a class-based validator.

In the .py config representation, the validator function can be passed directly as a callable.

#### Example configuration for external API key validation using HTTP request to Keycloak / OpenID Connect

This example shows how to validate API keys against an external service (e.g., Keycloak):
```toml
[api_key_check]
class = "lm_proxy.api_key_check.CheckAPIKeyWithRequest"
method = "POST"
url = "http://keycloak:8080/realms/master/protocol/openid-connect/userinfo"
response_as_user_info = true  # interpret response JSON as user info object for further processing / logging
use_cache = true  # requires installing cachetools if True: pip install cachetools
cache_ttl = 60  # Cache duration in seconds

[api_key_check.headers]
Authorization = "Bearer {api_key}"
```
#### Custom API Key Validation / Extending functionality

For more advanced authentication needs,
you can implement a custom validator function:
```python
# my_validators.py
def validate_api_key(api_key: str) -> str | None:
    """
    Validate an API key and return the group name if valid.
    
    Args:
        api_key: The API key to validate
        
    Returns:
        The name of the group if valid, None otherwise
    """
    if api_key == "secret-key":
        return "admin"
    elif api_key.startswith("user-"):
        return "users"
    return None
```

Then reference it in your config:
```toml
api_key_check = "my_validators.validate_api_key"
```
> **NOTE**
> In this case, the `api_keys` lists in groups are ignored, and the custom function is responsible for all validation logic.

## 🛠️ Advanced Usage<a id="-advanced-usage"></a>

### Dynamic Model Routing<a id="dynamic-model-routing"></a>

The routing section allows flexible pattern matching with wildcards:
```toml
[routing]
"gpt-4*" = "openai.gpt-4"           # Route gpt-4 requests to OpenAI GPT-4
"gpt-3.5*" = "openai.gpt-3.5-turbo" # Route gpt-3.5 requests to OpenAI
"claude*" = "anthropic.*"           # Pass model name as-is to Anthropic
"gemini*" = "google.*"              # Pass model name as-is to Google
"custom*" = "local.llama-7b"        # Map any "custom*" to a specific local model
"*" = "openai.gpt-3.5-turbo"        # Default fallback for unmatched models
```
Keys are model name patterns (with `*` wildcard support), and values are connection/model mappings.
Connection names reference those defined in the `[connections]` section.

### Load Balancing Example<a id="load-balancing-example"></a>

- [Simple load-balancer configuration](https://github.com/Nayjest/lm-proxy/blob/main/examples/load_balancer_config.py)  
  This example demonstrates how to set up a load balancer that randomly
distributes requests across multiple language model servers using the lm_proxy.

### Google Vertex AI Configuration Example<a id="google-vertex-ai-configuration-example"></a>

- [vertex-ai.toml](https://github.com/Nayjest/lm-proxy/blob/main/examples/vertex-ai.toml)
  This example demonstrates how to connect LM-Proxy to Google Gemini model via Vertex AI API

### Using Tokens from OIDC Provider as Virtual/Client API Keys<a id="using-tokens-from-oidc-provider-as-virtualclient-api-keys"></a>

You can configure LM-Proxy to validate tokens from OpenID Connect (OIDC) providers like Keycloak, Auth0, or Okta as API keys.

The following configuration validates Keycloak access tokens by calling the userinfo endpoint:
```toml
[api_key_check]
class = "lm_proxy.api_key_check.CheckAPIKeyWithRequest"
method = "POST"
url = "http://keycloak:8080/realms/master/protocol/openid-connect/userinfo"
response_as_user_info = true
use_cache = true
cache_ttl = 60

[api_key_check.headers]
Authorization = "Bearer {api_key}"
```

**Configuration Parameters:**

- `class` - The API key validation handler class ([lm_proxy.api_key_check.CheckAPIKeyWithRequest](https://github.com/Nayjest/lm-proxy/blob/main/lm_proxy/api_key_check/with_request.py))
- `method` - HTTP method for the validation request (typically `POST` or `GET`)
- `url` - The OIDC provider's userinfo endpoint URL
- `response_as_user_info` - Parse the response as user information for further usage in LM-Proxy (extend logged info, determine user group, etc.)
- `use_cache` - Enable caching of validation results (requires installing the `cachetools` package if enabled: `pip install cachetools`)
- `cache_ttl` - Cache time-to-live in seconds (reduces load on identity provider)
- `headers` - Dictionary of headers to send with the validation request

> **Note**: The `{api_key}` placeholder can be used in headers or in the URL. LM-Proxy substitutes it with the API key from the client to perform the check.

**Usage:**

Clients pass their OIDC access token as the API key when making requests to LM-Proxy.

## 🪝 Request Handlers (Middleware)<a id="-request-handlers--middleware"></a>

Handlers intercept and modify requests *before* they reach the upstream LLM provider. They enable cross-cutting concerns such as rate limiting, logging, auditing, and header manipulation.

Handlers are defined in the `before` list within the configuration file and execute sequentially in the order specified.

### Built-in Handlers

LM-Proxy includes several built-in handlers for common operational needs.

#### Rate Limiter

The `RateLimiter` protects upstream credentials and manages traffic load using a sliding window algorithm.

**Parameters:**

| Parameter | Type | Description |
|-----------|------|-------------|
| `max_requests` | int | Maximum number of requests allowed per window |
| `window_seconds` | int | Duration of the sliding window in seconds |
| `per` | string | Scope of the limit: `api_key`, `ip`, `connection`, `group`, or `global` |

**Configuration:**
```toml
[[before]]
class = "lm_proxy.handlers.RateLimiter"
max_requests = 10
window_seconds = 60
per = "api_key"

[[before]]
class = "lm_proxy.handlers.RateLimiter"
max_requests = 1000
window_seconds = 300
per = "global"
```

#### HTTP Headers Forwarder

The `HTTPHeadersForwarder` passes specific headers from incoming client requests to the upstream provider—useful for distributed tracing or tenant context propagation.

Sensitive headers (`Authorization`, `Host`, `Content-Length`) are stripped by default to prevent protocol corruption and credential leaks.
```toml
[[before]]
class = "lm_proxy.handlers.HTTPHeadersForwarder"
white_list_headers = ["x-trace-id", "x-correlation-id", "x-tenant-id"]
```
See also [HTTP Header Management](https://github.com/Nayjest/lm-proxy/blob/main/doc/http_headers.md).

### Custom Handlers

Extend functionality by implementing custom handlers in Python. A handler is any callable (function or class instance) that accepts a `RequestContext`.

#### Interface
```python
from lm_proxy.base_types import RequestContext

async def my_custom_handler(ctx: RequestContext) -> None:
    # Implementation here
    pass
```

#### Example: Audit Logger
```python
# my_extensions.py
import logging
from lm_proxy.base_types import RequestContext

class AuditLogger:
    def __init__(self, prefix: str = "AUDIT"):
        self.prefix = prefix

async def __call__(self, ctx: RequestContext) -> None:
        user = ctx.user_info.get("name", "anonymous")
        logging.info(f"[{self.prefix}] User '{user}' requested model '{ctx.model}'")
```

**Registration:**
```toml
[[before]]
class = "my_extensions.AuditLogger"
prefix = "SECURITY_AUDIT"
```

## 🧩 Add-on Components<a id="-add-on-components"></a>

### Database Connector<a id="database-connector"></a>

[lm-proxy-db-connector](https://github.com/nayjest/lm-proxy-db-connector) is a lightweight SQLAlchemy-based connector that enables LM-Proxy to work with relational databases including PostgreSQL, MySQL/MariaDB, SQLite, Oracle, Microsoft SQL Server, and many others.

**Key Features:**
- Configure database connections directly through LM-Proxy configuration
- Share database connections across components, extensions, and custom functions
- Built-in database logger for structured logging of AI request data

## 📚 Guides & Reference<a id="-guides--reference"></a>

For more detailed information, check out these articles:
- [HTTP Header Management](https://github.com/Nayjest/lm-proxy/blob/main/doc/http_headers.md)
- [Configuring fallbacks](https://github.com/Nayjest/lm-proxy/blob/main/doc/fallback.md)
- [HTTPS / TLS Setup](https://github.com/Nayjest/lm-proxy/blob/main/doc/https.md)

## 🚧 Known Limitations<a id="-known-limitations"></a>

- **Multiple generations (n > 1):** When proxying requests to Google or Anthropic APIs, only the first generation is returned. Multi-generation support is tracked in [#35](https://github.com/Nayjest/lm-proxy/issues/35).

- **Model listing with wildcards / forwarding actual model metadata:** The `/v1/models` endpoint does not query upstream providers to expand wildcard patterns (e.g., `gpt*`) or fetch model metadata. Only explicitly defined model names are listed [#36](https://github.com/Nayjest/lm-proxy/issues/36).

## 🔍 Debugging<a id="-debugging"></a>

### Debugging Mode<a id="debugging-mode"></a>

#### Overview
When **debugging mode** is enabled,
LM-Proxy provides detailed logging information to help diagnose issues:
- Stack traces for exceptions are shown in the console
- Logging level is set to DEBUG instead of INFO

> **Warning** ⚠️  
> Never enable debugging mode in production environments, as it may expose sensitive information to the application logs.

#### Enabling Debugging Mode
To enable debugging, set the `LM_PROXY_DEBUG` environment variable to a truthy value (e.g., "1", "true", "yes").
> **Tip** 💡  
> Environment variables can also be defined in a `.env` file.

Alternatively, you can enable or disable debugging via the command-line arguments:
- `--debug` to enable debugging
- `--no-debug` to disable debugging

> **Note** ℹ️   
> CLI arguments override environment variable settings.

### Printing LLM Requests and Responses<a id="printing-llm-requests-and-responses"></a>

The `print_stream` configuration option makes LM-Proxy print every LLM request
and response to stdout, with the response appearing token by token as it is generated:

```toml
print_stream = true
```

```text
Requesting LLM gpt-5:
    [User]:
        What is the capital of France?
LLM Response:
    The capital of France is Paris.
```

This works for both streaming and non-streaming client requests.

> **Warning** ⚠️  
> Intended for local development and troubleshooting only.
> Prompts and generated content are written to the console in plain text,
> and the output of requests processed in parallel is interleaved.

## 🤝 Contributing<a id="-contributing"></a>

Contributions are welcome! Please feel free to submit a Pull Request.

1. Fork the repository
2. Create your feature branch (`git checkout -b feature/amazing-feature`)
3. Commit your changes (`git commit -m 'Add some amazing feature'`)
4. Push to the branch (`git push origin feature/amazing-feature`)
5. Open a Pull Request

## 📄 License<a id="-license"></a>

This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.
© 2025–2026 [Vitalii Stepanenko](mailto:mail@vitaliy.in)

lm-proxy

File Explorer

Today's picks

Was this content helpful?

Recent Feedback

lm-proxy

File Explorer

Today's picks

Was this content helpful?

Recent Feedback

Download README